CyberGym-E2E - Patch-Only: leaderboard
CyberGym-E2E (2026) evaluates coding agents across the vulnerability lifecycle on real OSS-Fuzz vulnerabilities in C and C++ projects, with a 10-dollar cost cap and a 90-minute limit per task. This board is the patch-only success rate (%) on the 920-task benchmark, where the agent is given the known crashing input and crash log and must fix the vulnerability without breaking tests; higher is better.
Metric: Patch-only success rate (%) on the 920 OSS-Fuzz vulnerabilities (139 C/C++ projects) of the expanded benchmark: given the ground-truth PoC and crash log, the agent must produce a patch that removes the crash and keeps the functionality tests passing; with a $10 cost cap and a 90-minute limit per task; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 87.1 | #123 |
| 2 | Claude Opus 4.6 | 84.1 | #102 |
| 3 | Gemini 3.1 Pro (Preview) | 83 | #94 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=cybergym-e2e-patch-only · How It Works · Data refreshed daily, snapshot 2026-11-13.